This was tough, and took forever to get working but it seems to be good now. ### References - [Setup ROCm on Fedora](https://fedoraproject.org/wiki/SIGs/HC#Installation) - NOTE: [AMD Official Docs](https://rocm.docs.amd.com/en/latest/deploy/linux/index.html) did NOT work for me. - Setup a Python Virtual Environment - Install Python latest version () - Important: Cuda did not work for me unless I did [this]. - Export HSA_OVERRIDE_GFX_VERSION=10.3.0 - My GPU (RX 6950 XT) is LLVM Target gfx1030 --- ## Install ROCm on Fedora 38 ```bash # To enable access to GPU resources for non-root users, the users must be added to the video or render group. The AMD ROCm Installation Guide and FAQs recommend the video group for all ROCm-supported operating systems. # To add the current user to the video group sudo usermod -a -G video $LOGNAME # To check for ROCm support of the running HW, you can install and run `rocminfo`: sudo dnf install rocminfo rocminfo # OpenCL # ROCm OpenCL can be installed with: sudo dnf install rocm-opencl # As well, `rocm-clinfo` or `clinfo` can be installed to verify it is working, e.g.: sudo dnf install rocm-clinfo rocm-clinfo # HIP # HIP is included in Fedora 39 and is pending for Fedora 38. For Fedora 38, you will need to add --enable-repo=updates-testing to dnf if it's not available yet. sudo dnf install rocm-hip # Export HSA_OVERRIDE_GFX_VERSION # 6950 XT has LLVM Target gfx1030, hence set to 10.3.0 # Adjust for your GPU LLVM Target export HSA_OVERRIDE_GFX_VERSION=10.3.0 >> ~/.bashrc source ~/.bashrc ``` --- ## Install PyTorch ```bash # Install a Python virtual environment mkdir pytorch && cd pytorch/ python -m venv .venv source .venv/bin/activate # Install PyTorch # Preview (Nightly), Linux, Pip, Python, ROCm 5.6 pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/rocm5.6 ```